Electrical transmission system operators (TSOs) in several regions around the world are announcing new grid codes that require data centers to change how they respond to electrical fault conditions. These changes are designed to avoid the destabilizing effect of multiple large data centers dropping from the grid at the same time in response to grid power quality issues. This can exacerbate imbalances on the grid.

Incoming fault ride-through (FRT) rules aim to minimize the chance that data center actions will exacerbate the grid imbalance, by requiring data center electrical and mechanical systems to tolerate voltage and frequency disturbances of up to 15% for periods of up to several seconds. Data center operators will be required to change not only their disconnection controls, but also their reconnection process to the grid with the individual TSOs, making data centers active participants in grid management (see Power companies act to stop data center-induced blackouts).

Operators need to assess their existing UPS and battery systems as well as mechanical equipment, review their design standards and consult with their equipment suppliers to understand what internal actions are needed to meet these new requirements. As many new FRT requirements are still in the proposal stage, it is important that operators engage with TSOs and regulators to help ensure any required changes are reasonable and practical.

Practical response

Grid operators are adding two essential new requirements to grid codes to manage voltage and frequency disturbances: 

• Higher limits on voltage and frequency oscillations and a longer period during which data center systems must tolerate these disturbances before disconnecting from the grid.
• A requirement for operators to coordinate with the TSO on disconnecting the data center load from and reconnecting it to the grid. In some cases, individual TSOs are expected to establish a specific timetable for reconnection.

Fault ride-through

Uptime Intelligence has spoken with several vendors about the ability of their installed and new equipment to meet these new requirements. We looked at two categories: IT power supply (UPS and associated equipment) and cooling equipment.

Critical IT loads

Vendors have told us that most existing facilities can tolerate the extended FRT requirements with some combination of hardware, software and configuration changes. As manufacturers take steps to conform their products with the FRT requirements, operators should be able to outfit new facilities to meet the basic ride-through requirements.

The key question for UPS equipment is how much the input voltage and frequency can vary before the UPS can no longer deliver Class 1 voltage and frequency output (as defined by the IEC 62040-3 standard) to the IT equipment. Most manufacturers support a 10% voltage variance, and a few support 15%, with a maximum frequency variance of 3 Hz. Many installed UPS systems will require firmware updates to adjust the allowable input variation and the time before disconnection triggers a transfer to battery power. However, not all UPS systems can be reprogrammed or upgraded to meet FRT requirements.

Increasing the allowable input voltage and frequency deviation risks exceeding the allowable tolerances of the static bypass. If the UPS fails during a disturbance, power to the connected IT infrastructure may be lost. This risk needs joint analysis with the UPS manufacturer.

Increasing the allowable voltage variation risks higher current on the input side of the UPS system when a disturbance involves an under-voltage of 15%. To avoid interruption, facility operators need to ensure that circuit breakers and cabling to the UPS input can support the higher current.

During FRT conditions, the UPS system should disengage from "eco mode" (a common energy efficiency mode utilizing the bypass), as double conversion would be needed to maintain output power quality during a voltage and frequency event.

Overall, modifying the UPS system is not expected to cause any appreciable increase in wear on the system. Since these modifications would keep the UPS system on input power longer rather than on battery power, there would be no increase in wear on the battery system. Under standard operating mode, the IT infrastructure should be protected, as the UPS system will maintain the rated output voltage and frequency. However, if a UPS failure occurs or is in maintenance mode when FRT conditions arise, IT power supplies may be affected particularly by major frequency deviations, as most power supply units (PSUs) are typically rated to tolerate only 0.5 Hz deviation.

Mechanical systems

Mechanical (cooling) systems are harder to manage. Many mechanical systems are powered by variable frequency drives (VFDs), which convert AC to DC and back to AC at the required voltage and frequency. Depending on the manufacturer, model and age of the products, some systems will operate without issue, while others can trip because of their limited tolerance for the expanded voltage and frequency variations. Restarting mechanical systems after their shutdown may be manual, and ramping up to capacity will likely take several minutes, risking a thermal event in the data hall.

Older mechanical systems with across-the-line starters and motors are likely to struggle with the expanded tolerances. Even those that can tolerate the required voltage and frequency variations will likely wear out faster. Lower voltages increase current flow, placing more stress on motor windings. Lower frequency will reduce motor speed somewhat, putting stress on components driven by the motors.

Manufacturers advised Uptime Intelligence that the best approach may be to install a UPS system, with minimal battery response time and storage capacity, at the electrical feeds to the various cooling system components to maintain voltage and frequency. For new construction or major upgrades, flywheels or integrated diesel-rotary (DRUPS) systems offer an alternative approach to protecting mechanical loads.

Cooling and electrical system controllers are embedded IT systems with their own limited tolerance for voltage and frequency fluctuations and may be at risk of failure. The facilities team needs to determine the allowable variation and take appropriate action, such as installing a local UPS system, to protect critical controllers.

If the data center electrical system is configured properly, it may be feasible to ask the grid operator to exempt the mechanical loads from FRT and allow them to be transferred to generators while the IT load stays on the grid. In modern, low-PUE data centers, cooling and power distribution account for a small proportion of the total load (varying between 5% and 35%, depending on the time of year, for a facility with an average PUE of 1.3). Disconnecting this portion of the total load may be acceptable to the grid operator. This would require an electrical system designed to transfer some loads to generators while others stay on the grid. However, this is a technical possibility rather than a generally understood, let alone accepted, practice by TSOs.

Mandated time to return load

Part of the upcoming FRT requirements is the controlled reconnection, or restoration of full load, at the TSO's request following a disconnection or load reduction by the data center system. This is known as the post-fault active power recovery process (PFAPR). In some scenarios when the fault is within tolerance, this will mean the return of a disconnected data center load to the grid as soon as a fault has been cleared. This may pose another significant challenge. The Irish TSO, EirGrid, proposes requiring loads to return in half a second, while other proposals specify return times up to a few seconds.

However, UPS systems will not commonly switch from battery power that quickly; they must first verify that power quality is acceptable before slowly loading the rectifiers in a specified pattern. This slower transition was implemented to reduce the sudden, cascading UPS failures that occurred in some systems in the past. Operators will need to consult with the UPS manufacturer to determine whether the return time can be shortened.

The engine generator systems also check power quality before transferring the load back to the grid. If they use an open transition, it will likely take minutes to return the load incrementally once the grid is stable. A closed-transition return requires a longer transition time because the power from the engine generator system and the utility must first be matched, after which the load can be slowly transferred back to the grid. TSOs will want faster restoration of loads when bulk generation capacity has stayed online and recovered to full power.

Managing risks

Data center operators will need to assess their equipment's ability to establish how they can meet the new FRT and related PFAPR requirements, identify any necessary remedial measures, and update their design standards for new facilities. This will require a detailed engineering review of systems and engagement with equipment manufacturers to determine which firmware modifications, hardware upgrades or retrofits are required. As each facility will have its own hardware configurations and specific requirements from the local grid regulator or TSO, modifications and adjustments to design standards will be facility-specific.

Assuming the UPS system maintains its rated output voltage and frequency to the IT during the utility variance, the power supplied to the IT will remain stable. If the UPS fails to the static bypass during a voltage or frequency fluctuation, the IT will be exposed to the grid disturbance, increasing the risk of IT outage or PSU failure.

Cooling could be briefly affected during the transition to on-site power, but this is a typical operating scenario at most data centers, so it should be covered by the cooling system design and operating procedures. It is critical that the operations team can execute the required procedures correctly.

The situation in a colocation facility may be more complicated; any service level agreements (SLAs) between a colocation provider and its tenant must clearly define acceptable power and cooling performance and client notification procedures during grid events. Tenants ought to be made aware of how FRT requirements may affect risks to facility power and cooling and what measures the operator is taking to mitigate those risks. Some tenants may decide to make changes to the IT equipment within their racks (e.g., specifying different PSUs, incorporating rack-integrated battery-backup units for business-critical systems).

Engage with the process

TSOs and regulators are developing new processes to maintain grid stability that will require data centers to become active grid participants rather than passive consumers of power. EirGrid has been among the first to respond with new grid codes (see Ireland's new grid rules signal a shift in data center roles). In the US, the North American Electric Reliability Corporation (NERC) has published studies and proposals for regulations and standards directing TSOs to require data centers to extend their FRT capabilities and follow specific disconnection and reconnection processes during disturbances (see Draft grid rules position data centers as active grid participants). Other TSOs in Europe and around the world are processing similar proposals.

Grid operators and regulatory groups have told us that they want the data center industry to engage with them to help find a practical solution. Uptime Intelligence recommends that data center operators engage with them to explain what is possible with currently installed systems, the extent of the changes required by new ride-through scenarios, and the likely timelines for implementing any required changes to electrical and cooling systems. If operators are not involved in the development of regulations and standards, they risk facing unworkable or overly prescriptive requirements that are difficult, overly costly or impossible to implement.

It will also be important to establish conditional or temporary exemptions from the rules for existing facilities that pose a low risk to the grid but need an extended transition period to comply with the new requirements.

Failure to engage in the process or ignoring the resulting proposals will be counterproductive: any grid reliability event, even if short of a blackout, caused by the disconnection of a large data center load will put the industry in a difficult position at a time of rising public opposition to data centers.

Next steps

FRT requirements will make data centers more active participants in grid stability, but provisions will vary by facility, equipment configuration and local grid code. Operators that have sites in grid regions with new FRT requirements will want to begin with a site-specific engineering review: UPS systems, static bypass tolerances, batteries, generators, cooling equipment and controls.

Input from equipment suppliers is essential to define remediation options, costs and implementation schedules. The output of the review is an implementation plan that sets out funding and schedules for required firmware changes, revised settings, hardware upgrades and operating-procedure changes, with particular attention to protecting IT loads and achieving realistic reconnection times.

In parallel, it is important that data center operators engage early with TSOs and regulators to validate technical assumptions, agree on practical disconnection and reconnection procedures, and negotiate phased compliance plans or conditional temporary exemptions where immediate upgrades are not feasible. Early, evidence-based engagement will give operators the best opportunity to shape proportionate requirements and reduce operational, financial and reputational risk.

Note: The regulatory analysis provided in this Update is the opinion of Uptime Intelligence. Data center operators should validate the interpretations with their legal staff and any relevant regulatory authorities.